A detection method for granary storage states based on a support vector machine and domain knowledge

Automatic detection of granary storage states is an important technology to ensure national food security. In view of distribution characteristics of bottom pressure, this study theoretically explored a relationship between detected value of bottom pressure of granaries and granary storage states through the technologies, such as integrated learning and nonlinear regression. On this basis, this research built layout models of single-ring pressure sensors on the bottom of granaries and proposed a detection method for granary storage states based on domain knowledge of statistics of bottom pressure in granaries. The experiment demonstrates that the detection method for granary storage states based on a support vector machine (SVM) and domain knowledge shows high detection accuracy and low requirements for sensor performance and low detection cost. Therefore, it can meet the needs of remote online detection of granary storage states usually used.